Ataish Nehra

About Ataish Nehra

ML & AI Engineer with 4+ years of experience building NLP pipelines, custom RPA and predictive ML solutions with 2 years in LLM-powered systems, that drive automation, reduce costs, and improve decision-making. My work spans IT ticketing automation, fraud detection, compliance AI, and enterprise RAG pipelines, combining technical rigor with a strong focus on real-world impact. I specialize in- Fine-tuning LLMs & SLMs (Granite, Mistral, BERT) for enterprise use cases- Designing multi-agent & graph-based AI workflows (LangGraph, LangChain)- Building offline & cloud-native ML systems with Docker, Kubernetes, and cloud (AWS, Azure, GCP)- Delivering scalable solutions in Python, SQL, R, Spark, Kafka, and PostgreSQL Currently researching- Graph-based agentic AI (LangGraph-driven content visualization workflows)- Enterprise multi-agent systems for collaborative research and knowledge synthesis Testimonial (excerpt from CISO, Volta Inc.):“Ataish brought a level of expertise in AI that was both technically impressive and refreshingly accessible. He has a gift for breaking down complex ML concepts into understandable terms, earning trust through both intelligence and character. He is not only a brilliant AI specialist but also a team-oriented professional with a passion for innovation.” Education & Credentials: M.S. in Data Science, Northeastern University (Boston, MA)B.Tech in Data Science, Ajeenkya DY Patil University (Pune, India)Certified: AWS Solutions Architect, Azure AI Engineer, Tableau Desktop Specialist Community & Publications Published an paper: Social Media Depression Analyzer using Recurrent Neural Networks‍ Mentored peers and interns in LLM fine-tuning, ML pipelines, and NLP best practices at Volta & Halluminate Chief speaker at Volta Inc\'s AI Sales event(2025), presenting applied NLP & ML solutions for real-world business problems to 50+ C-suite level executives Personal mantra: Learning over knowing. Impact over buzzwords. Let’s connect Email: a••••••••@gmail.com | GitHub: https://github.com/AtaishNehra

Experience

1. Senior Machine Learning Engineer

   Paxcom India (P) Ltd - A Paymentus Company

   Oct 2025 — Present · Gurugram, IN

   Built a low-latency PII masking layer using Microsoft Presidio + custom regex in an LLM Gateway, achieving 100% redaction of emails/API keys with sub-1ms processing latency.• Improved prompt injection detection accuracy by replacing DeBERTa-based classifiers with SLMs (Llama-Guard-3-1B, ShieldGemma-2B, InjecGuard), reducing false positives (e.g, “Summarize”) while maintaining a 95.6% jailbreak block rate.• Designed a Semantic Fast-Path Router using BAAI/bge-small-en-v1.5 embeddings, reducing LLM inference cost by 80–90% and enabling <50ms latency for safe prompt routing.• Implemented canary-token-based system prompt leakage detection (OWASP LLM07) using UUID injection and deterministic matching, ensuring 100% accurate, zero-inference leakage prevention.• Engineered a fault-tolerant guardrail orchestration pipeline using Temporal with weighted consensus and HITL fallback, enabling fully auditable, resilient defense-in-depth LLM workflows.• Developed a post-generation safety validation layer using ProtectAI DeBERTa-v3 + InjecGuard to detect reassembled prompt injections, achieving 100% detection of adversarial outputs in test logs.

Education

• Northeastern University

  Master of Science - MS, Data Science

• Ajeenkya DY Patil University

  Bachelor of Technology - BTech, IT DS